I have a pipeline that reconstructs five objects but it is trained one object at a time. Each object gets good reconstruction at different iteration numbers. If i go beyond such iteraation the reconstruction get bad. I want to do some kind of early stopping. I have done one using validation loss but its not working as required.
